The bill for losing a job to AI does not arrive with the severance cheque. It arrives eleven months later, in the second year of a search that will not close — and Goldman Sachs has now put a number on the wait. Workers pushed out of AI-exposed jobs since 2025 needed about a month longer to land the next one, and were roughly 10 percentage points more likely to still be unemployed a full year after the layoff. That is not disruption on a chart. That is a sentence with a release date.
The finding landed in a client note published Thursday, written by Goldman economist Pierfrancesco Mei, and reported by Fortune. It matters less for what it says about the economy than for what it says about the person holding the envelope. One month of extra unemployment is a missed mortgage payment. Ten points is a coin-flip you now lose more often than you used to, and the coin keeps flipping for a year.
Pierfrancesco Mei — The Scar Forms Before the Recovery Does
The method is worth understanding, because it is the difference between a headline and an alibi. No government survey asks a laid-off worker whether AI took the job, so Goldman built the group itself from the Labor Department’s Current Population Survey, analysing more than 15,000 displaced workers who lost jobs between 2016 and 2025. Each occupation was scored for AI exposure, each industry for adoption, and each job for whether the position was eliminated outright. The top 30 percent on the combined measure were treated as likely AI-displaced — freshly released data covering 2023 to 2025 is where the first wave sits.
The control is the part that makes it hard to wave away. Before 2023, when adoption was still thin, Goldman found no meaningful gap between the two groups. The gap appears with adoption, not before it. Mei’s own framing is that the effect is comparable in size to what the bank measured during four previous decades of technological upheaval — and that the damage in those waves proved to be durable rather than temporary.
Durable is the polite word. In the earlier waves, workers who lost jobs to technology saw real earnings grow more slowly than other displaced workers for years afterward, and their odds of a second stretch of unemployment stayed stubbornly higher. The pattern took forty years to fully measure. It has had roughly two years in this one.
The 55-to-64 Cohort — 57.3 Percent Is Not a Recovery, It Is a Rejection Slip
Age is where the abstraction turns into a person. AI-displaced workers aged 35 and over took about five weeks longer than other displaced workers to get back into full-time work; the under-35 group took about three. Five weeks is not a statistic when you are 58 and the healthcare premium is due at the end of the month. It is a decision about which prescription you stop taking.
The government’s own counts, which cover all layoffs rather than only the AI-exposed ones, show the same tilt with more force. The Labor Department survey counted 7.4 million workers displaced between 2023 and 2025, up from 6.3 million in the previous survey window. Among those who had held their jobs at least three years, 72.9 percent of workers aged 25 to 54 were employed again by January 2026. For ages 55 to 64 the figure was 57.3 percent. Above 65, it was 38.6 percent.
Read those three numbers in sequence and the shape of it is obvious. The system reabsorbs the middle of the workforce and quietly loses the end of it. And the reabsorption is happening at a discount: of long-tenured workers who lost full-time jobs and found full-time work again, about 49 percent were earning the same or more, down from about 62 percent in the prior survey. The job came back. The salary did not come with it.
Andy Challenger — 120,136 Cuts and a Category Called Possibly AI
Now the count that sits underneath the analysis. Employers cited AI in 120,136 announced job cuts through September of this year — about 21 percent of all cuts, and the single most-cited reason — according to Challenger, Gray & Christmas. Since the firm began tracking AI as a distinct category in 2023, it has been named in 184,538 cut announcements. Those are announcements, not layoffs completed, and the gap between the two is where a lot of human misery gets booked.
The classification fight is real and it cuts both ways. Challenger tracked a Bronx hospital system, Montefiore, that eliminated 12 utilisation-review nursing positions after adopting outside software; the New York State Nurses Association filed a class-action grievance calling it AI replacement, while the hospital called that characterisation misleading. With no clear statement of the product involved, the cuts went into a bucket the firm labels technological update, possibly AI — 20,219 of those in 2025 alone. Twelve nurses do not move a national number. They do lose twelve paycheques while the accountants argue about the label.
The incentive runs one direction. Naming AI in a layoff announcement plays well with investors and badly with the people still on the payroll, which is precisely why the language has hardened into a boast. The messier truth, as Challenger’s own analysts concede, is that regulation will make companies vaguer, not clearer. The data will get worse exactly as the harm gets better documented.
Erik Brynjolfsson — The Door Shut on 22-to-25-Year-Olds First
There is a second front that the rehire numbers hide, and it is at the entrance rather than the exit. Stanford’s Digital Economy Lab, using payroll data, found in August that employment for workers aged 22 to 25 in the most AI-exposed occupations sits about 19 percent below where it would be had it tracked similarly aged workers in less-exposed jobs. The researchers, Erik Brynjolfsson among them, found the adjustment running through fewer hires rather than more firings — and no comparable gap for experienced workers.
Put the two findings side by side and the arithmetic gets ugly. The young are not being laid off; they are not being let in. The older are being let go and then held at the door for five extra weeks. A labour market can look calm in the aggregate while quietly rerouting two entire generations — you just have to count the right doors.
The Yale Budget Lab — Both Sets of Numbers Can Be Honest
None of this is settled, and the dissent deserves its line. The Yale Budget Lab, working the same federal survey data, has reported finding no sign that measures of AI exposure, automation or augmentation are related to changes in employment or unemployment. Challenger’s own chief workplace voice has argued that hiring is up and AI is reshaping the labour market without dismantling it. Read those alongside Goldman’s note and the honest conclusion is not that somebody is lying. It is that we are measuring an early-stage effect with instruments built for a late-stage one.
Mei says as much: it is still too early to judge the impact by education or by region. What he can already see is the shape of the next round. AI-displaced workers are more likely to leave their old occupation and industry entirely — and when they are rehired, a larger share — especially among older workers — land in roles carrying the same or higher AI exposure. They did not escape the technology. They were reseated in front of it.
That is the detail the quarterly earnings call never mentions. Reskilling into a role that the same technology is aimed at is not a rescue; it is a deferral with better paperwork.
What the Next Twelve Months Will Actually Cost
The number to watch is not the layoff count. It is the duration — how long the gap runs for the people already out, and whether the extra month becomes two. Duration is where savings turn into insolvency, where a career becomes a series of contracts, where a 58-year-old stops applying because the silence has become an answer.
Goldman’s note is a warning dressed as research, and the warning is specific: the costs of displacement are larger in a downturn, and AI-related losses are expected to accelerate if the economy slows. Which means the experiment is about to be run a second time, in worse conditions, on people who have not yet recovered from the first pass.
The technology is not the villain here, and it is not the saviour either. The villain is an accounting habit that books a layoff as a saving in the quarter it happens and never books the twelve months of searching that follow. Somebody pays that. It is just never the person who signed the memo.
Sources: Fortune, the Bureau of Labor Statistics, Challenger, Gray & Christmas, Stanford Digital Economy Lab.